12 core / 24 threds are out ! Xeon E5-2697 v2

[t.petrov]

Hm, so you're comparing 12 threads vs 12 threads, this explains the minimal difference. If you compare them 12 vs 24, the xeons should be close to twice as fast.
Hopefully we'll be able to fix you're Yeti issue and then you'll be able to utilize the full power of the Xeons

[Morbid Angel]

Hm, so you're comparing 12 threads vs 12 threads, this explains the minimal difference.

No not correct. 12 threads in i7 is 6 physical cores. 12 threads in dual six core xeon is 12 physical cores. In fact, in my comparison I've enabled HT, on the xeon, just to be on the same side of things, and that's where the xeon performed slightly faster (what I've stated). Without HT the dual six core was the same speed as the single 6 core i7.

[waaazoo]

Dmitry,

I would say that there is either something wrong in your rig with HT or something wrong with Vray (too much threads ?). It's quite strange to have only a 15% gain when enabling HT compared to when computers are similar without HT (for the Xeon of course)
what a frenglish !

[Morbid Angel]

I don't know man. Its not just one rig, I have several machines in the office and several at home, which are all in similar ballpark of setup, xeons on supermicro boards vs i7 on asus boards. I this I will do some more renders and post results here.

[t.petrov]

If you have scenes that don't scale well on the Xeons you can send us, so we can check them.

[Morbid Angel]

If you have scenes that don't scale well on the Xeons you can send us, so we can check them.

I think what I'm going to do is core per core tests fist, then take it from there. In all of my productions since I've started using i7's, they outperformed xeons. But then again, all of my productions always have stuff like hair, vector displacements, multiple dynamic textures etc

[3LP]

Mate, just use cinebench, it's been my ultimate choice to compare cpu, and they are actually used in nearly all the bench test that geek sites are publishing, so it's easy to track...

My i7 3930k OC a 4.2 or 4 (don't remember) at home perform a bit more than 12 in cinebench.

http://www.cbscores.com/index.php?sort=rend&order=desc

Here is a list of cpu tested through cinebench, but there are plenty of charts out there, just flick through and you will have a good idea of what your cpu and the new cpu are worth in term of pure rendering power...

Edit :
Here is one of the bench done by tomshardware :

http://www.tomshardware.com/reviews/...ks,3585-6.html

So it seems to be 17.63 in score at stock clock, I don't know how they will overclock and if there are already motherboard supporting OC for those, but 17.63 compared to my 12 means that it is 46% quicker.
Ok fair enough, but knowing that only one cpu is worth $2949.69 (and you need 2 to match those 17.63 in cinebench) means that you're already paying $5899.38 only for the cpu, no motherboard no ssd, no anything, so the whole setup will be close about $8000... No?

Well, with the i7 3930k, a 32Go node will be purchased at $1500. so two of those will outperform your dual cpu xeon quite a lot, for only $3k...

Just my 2 cents

Stan

[t.petrov]

(and you need 2 to match those 17.63 in cinebench)

As far as I can see the 12 core xeon does 17.63 for a single socket machine, a two socket machine will go close or over 30.

[Morbid Angel]

As far as I can see the 12 core xeon does 17.63 for a single socket machine, a two socket machine will go close or over 30.

single 12 core? its still worse core per core isnt it? I've done some quick tests over here, in some cases i7 performs just as well as my xeon counter part, in others xeon is faster, I think it really depends on a situation, reason I want to do these tests with vray is that, after all we are rendering with it, not with cinebench, so their results are not really relevant for our purpose (IMO).

[ivaylo.ivanov]

More cores means more power per socket and less CPU frequency. You will not get 2x3930k or 3x3770K, the only benefit is that you are getting similar performance using one machine. Cinebench and similar are good just to compare result from reviews and check if everyting is working as expected.